Context

The UrBEn-ID (Urban British English and Identity) project was a two-year study (July 2014–July 2016) funded by The Leverhulme Trust, which explored the use of language and other semiotic practices in the enactment of identities among fourteen to sixteen year olds in inner city Manchester. The ethnographic part of the study took place in the academic year 2014–15. The project's two main research sites were inner-city learning centres within the Manchester Secondary Pupil Referral Unit (PRU), which cater for years ten and eleven (aged fourteen to sixteen) students who have been permanently excluded from mainstream education. The young people must attend every day during normal school hours (9.00–3.00), and study a reduced curriculum of core subjects for GCSE.Footnote ² By definition, all of the pupils in the learning centres have issues of one kind or another with the school context, hence they have ended up where they are—separated from mainstream with very little chance of returning.Footnote ³ The reasons for their exclusion varied widely—from fights, to bullying, to confrontations or aggression towards teachers and school staff, to general and persistent more minor discipline issues, often brought about by an inability to adapt to the mainstream environment. Many times we never knew the reasons for exclusion as we never asked; we just listened if they wanted to tell us. In addition (and perhaps not unrelatedly) to their volatile experiences with the school system, many of the young people had unstable family lives. Again, we did not ask, but the information we gathered either from the young people themselves or from the staff suggested a variety of home life experiences ranging from difficult to traumatic.

The centres are staffed by a variety of people. Each has two coordinators who are in charge of the day-to-day running of the centre and who usually have some kind of youth work background. In addition to being in charge of the centre, the coordinators sit in on classes as additional support for the subject teacher. The subject teachers work on a peripatetic basis—moving between centres to deliver classes in the core curriculum subjects of English, math, science, and art. There are additional classes in Preparation for Working Life (PfWL) and Personal Social and Health Education (PSHE), the first of which is delivered by a specialist teacher, the second by one of the coordinators. The final key staff member is the youth worker—a permanent member of the staff in each centre who offers general support in and out of classes. Due to the small and enclosed nature of the environment, there is a real sense of everybody pitching in. Meals are prepared by whichever staff members are keen and capable, be that coordinators, youth workers, or even caretakers, and staff roles seem very fluid, at least to an outsider. In between classes, pupils are generally free to play pool, table tennis, watch TV (usually music videos), or go outside and smoke. At one of the centres they are also free to go to the local shop, although they are supposed to let staff know they are going. At the other centre the shop is further away and they are not allowed to go there.

The result of all of this is a lively, often very noisy environment in which young people are free to express emotions, disagreements, and their own individual personalities within the enforced boundaries of fairness and awareness of others. Arguments (often extremely intense) are frequent, although physical fights are rare. Battles between staff and students are everyday occurrences, often involving the same individuals. Attendance was patchy, although often predictable on an individual basis, with some rarely missing a day, and others turning up only occasionally.

The participants

The twenty-five participants discussed here all attended one of the two centres over the course of our time there. Some were present throughout the period, others were more transitory. Some were previously known to me due to a pilot study I carried out at the same locations the previous year. Table 1 provides pseudonyms, sex, centre, year group, and ethnicities of the participants. Ethnicities were established through a combination of self-reporting and staff-reporting. Social class was never a focus of the project, as the participants represent a uniform sample in terms of social background and educational expectations and attainment. This is not to say that class is not relevant, simply that there was no meaningful social variation, meaning that the study is, by default, one of this particular socioeconomic group. The PRU population in general tends to represent the lower end of the socioeconomic scale, with pupils known to be eligible for and claiming free school meals (indicating low-income families) being around four times more likely to receive a permanent or fixed-term exclusion than those who are not eligible (Department for Education 2016). As far as I know, all participants were born and raised in the UK, the vast majority in Manchester.

Table 1. Participants.

Quantitative analysis

The data presented here represent a mixture of recording contexts; where possible, however, each speaker's total includes at least some data from a ‘research meeting’—a conversation/interview with one of the two researchers in which the young people were either alone or in pairs. The purpose of this was to try to ensure at least some sense of comparability, albeit in a limited way. Data collected from the mock college interviews were excluded on the basis of this being a very different context in which only a few participants were involved. Not all of the data we recorded are suitable for this type of phonetic analysis, so what is presented here represents a large sample of the total. As the analysis progresses throughout this section, the data set is made more specific and is reduced. Table 2 provides a contextualising overview by showing the results of the auditory analysis of all 886 tokens of (th) from the speech of twenty-five participants. This represents a mean rate of just over thirty-four tokens per speaker; however, there is some variation in the total for each individual, mainly due to the methodological issues mentioned above (attendance, cooperation, etc.).

Table 2. Overall results of the auditory analysis for (th).

Preliminary auditory analysis of the (th) variable was carried out at the time of transcription of the audio recordings using Elan. The transcriber was one of the two researchers or one of five employed transcribers; they highlighted some of the more obvious variants as they occurred. More careful auditory analysis was then carried out by me to check those tokens already identified, and to identify all other instances of the variable. Tokens that were unclear even after repeated listening due to poor recording quality or overlapping speech were marked as such and discarded from further analysis. The only other issue that emerged in the process was the difficulty in discriminating between [f] and [θ] on occasions. Sometimes it was clearly one or the other, but other times the two were almost indistinguishable. Here the field notes and prolonged observation helped, as both researchers had noted several times in our day-to-day observations the almost complete absence of [θ] in the speech of any of the participants (the difference is, of course, much easier to distinguish with additional visual cues). As a result, I was predisposed to identifying genuinely in-between tokens as [f]. This would be potentially problematic in a study of th-fronting specifically, but as the main focus here is on th-stopping, I feel satisfied with the process.

All speakers

Figure 1 shows the distribution of the 819 [θ, f, t] (th) tokens among twenty-five speakers, ordered from the right by frequency of [t]. [h] and [ʔ] tokens have been removed due to very small numbers and limited obvious relevance to the analysis, as have tokens that were unclear. Two things immediately stand out from the chart: (i) that fronted [f] is by far the most frequent variant among the group as a whole, accounting for almost 85% of the total and 100% of the tokens for five of the speakers; and (ii) use of [t] is very limited, both in terms of overall frequency (5% of the total) and in terms of distribution among speakers (eight of twenty-five).

Figure 1. Distribution of the 819 [θ, f, t] (th) tokens, ordered from the right by frequency of [t] (twenty-five speakers).

Multiple logistic regression analyses were carried out using Rbrul (Johnson Reference Johnson2016), including individual speaker as a random effect. Rbrul is a variable rule program in the mould of Goldvarb (Sankoff, Tagliamonte, & Smith Reference Sankoff, Tagliamonte and Smith2005), yet which incorporates mixed-effects modelling. The result is a model which ‘can still capture external effects, but only when they are strong enough to rise above the inter-speaker variation’ (Johnson Reference Johnson2009:365). Rbrul expresses coefficients in log-odds rather than factor weights, although both are given in the analysis presented here. The first analysis was carried out on the data presented in Figure 1, except that those speakers with fewer than ten tokens (Danielle, Jenson, Megan) were excluded. These exclusions resulted in 800 tokens from twenty-two speakers being entered into the model, with thirty-five realisations of [t]. The dependent variable was (th) with the application value as [t] and the nonapplication values as [f] and [θ]. The independent variables are described in Table 3. Each independent variable was included based on the plausibility of it playing some kind of role in the observed variation, informed by previous studies or by experience of the context. The results of the analysis can be seen in Table 4. It should be pointed out that two of the tokens in particular are ones that might normally be excluded from traditional quantitative variationist analysis on the basis of their context of use not being natural to the speaker. Both were used in the process of describing someone else's speech, so in that sense they did not ‘belong’ to the speaker using them. However, this highlights a problem with traditional variationist research, as it begs the question of where to draw the line. Yes, these examples are clearly not occurring as a natural part of these speakers’ repertoires (at least not in these instances), but excluding them necessitates the implementation of a cut-off point for other less clear examples. Rather than get into a situation whereby the researcher decides what is and is not ‘natural’ for a particular individual, I am taking the approach that all language is data, and if it occurred, it is part of the linguistically constructed reality of that moment.

Table 3. Independent variables for the regression analysis.

Table 4. Regression analysis (Rbrul) of the effect of linguistic and social factors on the realisation of voiceless (th)—twenty-two speakers with > ten tokens.

Before looking at the statistically significant variables, it is perhaps worth looking, in particular, at one of those that did not reach statistical significance in this model: ethnicity. Ethnicity is of particular interest here, as it of course relates to points made earlier with regard to the links between th-stopping and black varieties of English. However, ethnicity does not emerge as a statistically significant influence on the realisation of /θ/ as [t]. It would be easy to simply state this fact and move on; however, there is clearly a correlation between ethnicity and use of [t], even if it is not strong enough to emerge as significant among the other variables. As can be seen in Table 5, the percentage of [t] realisations among white British speakers is considerably lower than almost all the other ethnic groups, and if we were to take a very blunt ‘white’ vs. ‘not white’ approach, the difference would be very large indeed. While this approach is fairly common in variationist studies, it rarely explains anything in itself. This is not to say there is no value in it at all—looking for correlations such as these then opens the door to investigating why they might be occurring, but in themselves they do not carry much meaning (cf. Cameron's Reference Cameron, Joseph and Taylor1990 ‘correlation fallacy’). After all, ethnicity itself is such a fluid concept that is constructed in practice rather than something passively inhabited (e.g. Schilling-Estes Reference Schilling-Estes2004), so unless one looks at how it is being constructed, a simple measurement offers limited insights. In addition, straightforward correlations (and statistical analyses such as the ones above) miss the important fact that in this particular case, three White British speakers did use [t], and two Mixed White/Black Caribbean or Black Caribbean speakers did not. Admittedly, when we are talking about such large data sets as can be found in a lot of first wave variationist studies (e.g. Labov Reference Labov1972) in which such correlations are central, these kinds of outliers are perhaps of little relevance; after all, the aims of such studies, especially in relation to issues of identity, tend to be very different (see Drummond & Schleef Reference Drummond, Schleef and Preece2016 for a discussion). But in a smaller-scale study such as this, a single instance of a variant can, arguably, provide a lot of information: ‘Indexical meaning can… arise out of statistical commonality or single instances of use that are salient enough to gain meaning for speakers’ (Kiesling Reference Kiesling2009:177). But this observation works both ways—on the one hand, the existence of these outliers warns us against drawing blunt ethnicity-related conclusions, but, on the other, we do still need to account for the fact that the rest of the speakers did fall into such a pattern. One possibility is that ethnicity is playing a more indirect role, perhaps mediated by another social factor (or combination of factors). This idea is explored shortly.

Table 5. Use of [t, f, θ] divided by speaker ethnicity. Percentages show proportion of total tokens for that group.

With regard to those factors that did emerge as statistically significant, the models suggest that informal contexts are more likely to generate realisations of [t] than formal contexts (research meetings). In fact, only four of the thirty-five [t] tokens appeared in a formal context. While this is perhaps not surprising, it is of some interest in that it reinforces the observed similarities between the various informal contexts. For example, this category groups together the lessons themselves and the breaks between lessons, two situations that might usually be seen as something quite different in a ‘normal’ school environment. But this is not a ‘normal’ school environment—classes are very small (could be anything from one to seven pupils) and the atmosphere is very relaxed. So-called banter between students (see Dray Reference Dray2017) is as common in the classroom as it is outside, and while they may be pulled up for excessive swearing, for example, in class, the same can happen outside class. However, when we invited individuals or pairs to come and talk to us separately, however informal we made it, this was still an unfamiliar and relatively new situation for them, and one that tended to inhibit use of [t]. The possible reasons for this are explored in the next section.

The raps factor is of interest as it begins to identify a possible social practice aspect of the variation rather than a macro-social influence. All it suggests at this stage is that those individuals who we observed rapping as part of their everyday practice (Abdou, Aiden, Callum, Georgia, Jacob, Jamal) are more likely to use a [t] variant. Crucially, however, only two of the recorded tokens of [t] being discussed here occurred during rapping; the rest occurred as part of ‘normal’ speech. Position is of passing interest, but is largely predictable, given that thirty of the thirty-five instances of [t] occur in word initial position. It is also interesting to note that of the six individuals who we observed rapping, three are white British, one is Black African, one is mixed white British/black Caribbean, and one is mixed white British/Pakistani. The reason this is important is that in a study with a larger sample it might be expected that the traditional association of rap/grime with black culture results in a greater likelihood of ‘rappers’ being nonwhite. In this case, ethnicity might emerge as a statistically significant factor alongside rapping, or else the argument could be made that ethnicity is being mediated by rap, thus still playing a role. I cannot discount this interpretation, and it potentially brings in useful discussion around what resources are available to be used by, and how ethnicity can be enacted differently within, different groups of people. However, the data that is available here do not allow that investigation in any meaningful way.

With raps emerging as significant, it might seem strange that music does not do the same. However, I think this is more to do with the way in which we coded the data for musical tastes. If we had our time with the young people again we would delve more deeply into this, but the reality is that our knowledge is limited and patchy in this area. The categories we used are vague and overlapping, and we did not manage to gather information consistently. Rapping is far more robust, as this is a straightforward measure of an observed activity.

A further model was tested including only the nine individuals who had produced [t] (342 tokens in all) and this time the only two factors emerging as statistically significant were again formality and position (see Table 6).

Table 6. Regression analysis (Rbrul) of the effect of linguistic and social factors on the realisation of voiceless (th)—nine speakers who produced [t].

One final point should be made in relation to the more quantitative aspect of the study—the fact that [t] appeared only in a very limited set of lexical items (thing x 23, thief x 5; everything x 2; three x 2; birthday x 1; teeth x 1; anything x 1). Although lexical context was excluded from the model on the basis of the distribution of the variants being so skewed between the items, it is of course of interest that only certain words appear able to use [t]. Clearly the most frequent context is thing, which is perhaps no surprise when we consider the grime lyrics from earlier. In fact, it would be plausible, and in many ways more robust from a quantitative perspective, for this whole study simply to focus on variation within thing. Footnote ⁴ With that in mind, one final model was tested that included only thing related words (thing, anything, everything) that were realised as [θ, f, t]. Individuals with fewer than five tokens were excluded. The results can be seen in Table 7. Interestingly, the only factor that emerges as statistically significant this time is raps.

Table 7. Regression analysis (Rbrul) of the effect of linguistic and social factors on the realisation of voiceless (th)—only thing words.

Clearly none of the models can fully explain the situation, as each is necessarily partial and limited in its own way; with such small numbers (thirty-five instances of [t] overall) and with such skewed data, we should be cautious about reading too much into them. Also, by inputting raps as an influencing factor, we are in danger of simply substituting ‘practice’ categories for macro-social categories. Taken as a whole, however, the results above do serve a purpose in that they provide insights into some possible patterns, highlighting areas that perhaps warrant further investigation. Clearly there is something about both being in an informal context, and being someone who raps, that predisposes an individual to produce [t], even taking into account the limited lexical range in which such variation is possible. In addition, it would appear that straightforward ethnicity is not a relevant explanatory category to pursue. But at the moment these remain abstract and somewhat meaningless correlations. In order to gain insights into what these patterns might mean, we turn to the interactional data below, which highlights two specific contexts.

Interactional analysis

One context that plays a significant role in the data here is the weekly art lessons in one of the centres. These are generally very laid-back sessions, lasting longer than normal lessons (over two hours), with a teacher who enjoys engaging with the class in casual conversation as they are working, and an extremely experienced youth worker, Michael, who is well-respected and who relates to the young people especially well. The class contains four of the [t] producers, and the context (over various recordings) accounts for nineteen of the thirty-five tokens of [t].

The four individuals under discussion could loosely be described as friends within the context. Jamal is undoubtedly the ‘leader’, as is evident from the behaviour of the others around him. Casually but immaculately dressed in either sweatpants or jeans and t-shirts, Jamal is slim and quite small compared to the others, although he carries himself with apparent confidence, at least to an observer. Jacob is a sportsman—he is well-built and physically fit, and often talks of his physical/sporting activities. He is viewed by the others as a bit of a joker,Footnote ⁵ and dresses casually in sweatpants and t-shirts. Abdou, throughout our time in the centres, appeared to constantly seek Jamal's approval, with Jamal in turn mocking him at almost every opportunity. Often the target of the mocking was Abdou's appearance, with Jamal regularly describing him as “ug” (‘ugly’). Abdou seems to put up with a lot of abuse (mostly verbal but occasionally physical) but keeps coming back for more. Jake is the outsider of the group—coming from a different part of Manchester, he would rarely see the other boys outside school. He is bigger than the others, but not toned like Jacob.

Topics of conversation in the art class tend to be quite wide-ranging; however, a recurring theme is one of ‘toughness’, be that fights the boys have either had or witnessed, getting into trouble with the police, or demonstrating how friendly they are with known criminals or possible gang members. Extract (1) took place during a conversation in which all four boys were present, along with Michael, the youth worker (aged around fifty). Also in the room were one of the girls, Shanique, one of the centre coordinators, Joy, and me. However, Michael and the four boys were sitting together, with everybody else outside of this small group. They were working on mosaics—which at this moment meant arranging cut-up images onto a page. The conversation leading up to the extract was about somebody (let's call him Marcus) they all knew who was in prison, and when he was getting out. Here, they are discussing when they last saw him.Footnote ⁶

1. (1)

   

Abdou is the first to use a [t] variant in line 20 when he is talking about Marcus being a “tief”. The reason this is of interest is because up until this point in the recording, Abdou had produced nineteen tokens of voiceless /θ/, sixteen of which were realised as the ubiquitous [f], and three as [θ]. Incidentally, all three [θ] tokens occurred while he was singing as he worked, so it is likely that he was simply replicating the standard variant within the original version of the song itself. Jacob produces a similar realisation of the same word in line 22, having previously produced eleven tokens of voiceless (th) comprising seven [f], three [θ] and one [t]. So what is it that brings about Abdou's use of [t] at this particular point?

I think there are two plausible (related) explanations. Firstly, from a stance-based perspective, Abdou and the others are clearly setting themselves up as being part of this tough, criminal life. There has already been talk of when each person last saw Marcus, thus establishing their credentials as being close to that world. They are not necessarily trying to outdo each other, but they are each trying to lay claim to at least being in close proximity to Marcus. Abdou and Jamal are slightly ahead here given the fact that Marcus was actually with them on the bus (line 1), and they were all part of whatever misdemeanour is mentioned by Jamal (unfortunately unclear) which led them all to “run for it” (lines 5–7). Jamal then lays special claim to being close to Marcus by emphasising that Marcus used to chill with him and not Lamar (line 15). Abdou then delights in recalling when Marcus robbed “Lamar's” (presumably Lamar's house) which is when he says that he used to be a “tief”, a sentiment then echoed by Jacob. It could be argued that the use of [t] is one element of the style Abdou is creating and manipulating in the context of ‘tough-talk’. In using [t] in this environment, Abdou is reinforcing the stance he is taking in which he is trying to align himself alongside Marcus as an insider to this world. He has already provided his credentials as being a friend of Marcus, and now he is reminiscing (in a very positive way—there is laughter in his voice) about his criminality. In order for this interpretation to work, there must of course be something indexically linking [t] to a stance of toughness. I return to this shortly.

An alternative explanation calls on lexical rather than phonetic variation. The fact is, of the five examples of the word thief we have analysed in the recordings, all are realised with the [t] variant. Although impossible to say for sure with such a small sample, there is a possibility that thief and tief are seen as two distinct words with slight differences, if not in referential meaning, then at least in contextual appropriateness. The difference is subtle, but such an interpretation renders this less an instance of th-stopping and more one of lexical choice. This interpretation is supported by the fact that in line 23 Michael, the adult youth worker, also uses the [t] variant. Although this is specifically not a study of the adults’ speech, the language of the adults as interlocutors is of course relevant, as it forms part of the interactionally created reality. There is no reason to suspect that Michael was aligning himself alongside Abdou and the others in terms of their positive stance towards the world of criminality; rather, his use of tief was a predictable lexical choice given his own Jamaican heritage and very noticeable Jamaican English dialect and accent, an accent that very much includes variants of /θ/ as [θ, t] (Wells Reference Wells1982:565) (although not [f], which indeed is a variant I did not hear Michael use). In this sense, tief simply is the Jamaican English word for thief, so this is the natural word for Michael to use. For the boys, however, there is no such natural explanation; the boys do not have Jamaican English dialects.

There is a possibility that the increased use of [t] in this context comes about as a result of the boys accommodating towards Michael's speech. I cannot dismiss this interpretation, as Michael is present in the majority of the art lessons (the situation in which most of the recorded th-stopping occurs). If this were the case, however, I would have expected a lot more th-stopping to emerge in other situations, as Michael is with the boys most of the time in one way or another. But this does not appear to happen, either in the recorded or observed data. The interpretation is therefore not impossible, and it might be the case that individual instances can be reliably interpreted as coming about in this way, but it is an unlikely explanation of the overall pattern. There is also a strong possibility that there is a priming effect in operation—that one use of [t] in a particular context encourages a repeated use. Although this is likely, however, I would argue that it does not detract from the stance-taking interpretation. It may lower the bar for the availability of [t] to be used, but its use is certainly not inevitable, and this would not explain its initial use.

The boys are choosing (for want of a better word, see earlier discussion surrounding the conscious/unconscious distinction) to use this particular variant. At which point, the distinction between the two explanations (lexical or phonetic) largely ceases to be relevant, as the question simply becomes—what is the indexical link between a stance of toughness and either tief or [t] in the speech of the boys? I believe one possibility lies in the cultural associations of th-stopping in relation to music. Recall that raps emerged as statistically significant in the model, suggesting that those individuals who we observed rapping were more likely to produce [t]. In fact, three of the four the boys here (Abdou, Jamal, Jacob) and both of the th-stoppers are ‘rappers’. But we are talking about a specific type of rap—that practiced in UK grime and Jamaican dancehall. Grime especially is very down to earth, and very much of the street. It is the story of hardship, of disenfranchisement, of the daily battles faced by what many see as the UK's underclass (MacDonald Reference MacDonald2008). Grime belongs to the poverty-stricken inner city, and while it crosses racial divides, it is socially grounded in a particular social class (Dedman Reference Dedman2011:519). Grime is tough. And most importantly, grime speaks to urban British youth in the language of urban British youth, and its participatory nature thus sets up a cycle of linguistic reinforcement. Grime (and, to an extent, dancehall) is not a type of music that people passively listen to, it is a type of music they engage with (Dedman Reference Dedman2011, Drummond Reference Drummond2016). Its history is one of young men creating homemade tapes of themselves rapping, and this DIY approach has continued; the tapes have simply been replaced by phone videos and YouTube. The fact that we observed these young men rapping in our day-to-day attendance at the learning centres is crucial; it helps to demonstrate how embedded the music is in everyday life. We have several examples of boys (including Jamal and Abdou) seamlessly slipping into rap as part of a conversational interaction, as if the lines between their ‘normal’ conversational speech and their ‘grime’ speech are blurred or even nonexistent.

I am wary of overstating the specific grime significance here, as many of the characteristics and social connections described above are equally relevant to hip hop more widely (e.g. Cutler Reference Cutler1999; Alim Reference Alim2006; Alim, Ibrahim, & Pennycook Reference Alim, Ibrahim and Pennycook2009; Quinn Reference Quinn2010; Jeffries Reference Jeffries2011). However, I would suggest that the especially immersive nature of grime sets the findings from this study apart from apparently similar studies such as Cutler (Reference Cutler1999), which looks in great detail at ‘Mike’ and his use of AAVE features partly as a result of his alignment with hip hop (see also Bucholtz Reference Bucholtz2011). While there are many comparisons to be drawn, especially in relation to the lifestyle Mike became part of, I would argue that listening to, and learning lyrics of, well known hip-hop tracks (as Mike did) is subtly yet crucially different from actually engaging with the production of grime.

In summary, what I am arguing here is that the boys are using this particular variant in the process of adopting a stance of toughness in relation to the interaction that is taking place. The variant acquires this social meaning in this particular context through its association with grime (and rap/hip hop and dancehall more generally), which embodies a street/urban/tough lifestyle. The quantitative analysis gave a partial picture by highlighting the possible connection between use of [t] and ‘rapping’, and this then formed a basis for the interactional analysis.

However, it is not always the case that [t] is used so explicitly to align with a tough stance in relation to the content of the particular interaction taking place. It can also be used more generally to index participation (or desired participation) in that world of grime, especially when such involvement does not perhaps come quite so naturally. Callum is very different to Jamal and his friends. At the other centre, he is an energetic and generally cheerful boy who, although seemingly confident, does not appear to have the swagger or coolness of Jamal especially. Callum is white, interested in grime, and we have observed him rap. In the thirty-four tokens of /θ/ we have recorded of him, twenty-eight are [f], four are [t] and two are [θ]. Two of the instances of [t] come close together in a recording made, again, during an art lesson. Interestingly, grime had itself been the topic of conversation moments before the extracts below, with Callum actually rapping in class and then talking about one of his favourite artists. In extract (2), Callum is wandering around the room, and seems to be looking for the voice recorder.

1. (2) Callum:  yo where's that little recorder thing <[tɪŋ]>?

When he finds it he says:

1. (3) Callum:  yo this th-<[f]>- this thing's <[tɪŋz]> still on.

Extract (3) is very telling, as Callum self-corrects from fing to ting, an apparently conscious move. This correction arguably illustrates the role of intent discussed earlier; neither of the identities that Callum performs before and after the self-correction of this contextually salient variable is any less real than the other, yet the second one is the one he intended to perform at that particular moment. By consciously using a particular variant, Callum is taking a stance in an attempt to (re)align himself as someone involved in grime and all it represents. Without that self-correction, we are glimpsing another side of Callum—the side that he likely performs in other contexts, and, arguably, the side that comes most naturally to him given the social constraints of his particular life experience. It is almost as if for a second he forgot the identity he was performing at that moment, and consciously re-adopted his preferred stance in relation to the grime lifestyle. Perhaps [t] is an example of a feature that at the moment remains conscious and awkwardly intentional, yet one that, in time, will become automatic and more integrated into his own style (Eckert Reference Eckert and Coupland2016:78). This scenario brings to mind an incident described in Cutler (Reference Cutler1999:429) when Mike, a white teenager who uses AAVE speech features as part of his increasing identification with hip-hop culture, self-corrects his standard pronunciation of [æsk] to the more desirable AAVE version of [æks]. A similar occurrence is described in Bucholtz (Reference Bucholtz2011:34) when she suspects a young male speaker is ‘correcting away from Standard English’ in his use of remote past been. Arguably, Callum's need to perform this identity at this point was made especially acute due to an incident that had occurred a few minutes previously (extract (4)), just after he had been rapping. Having heard him rap, the teacher (James) asked him what it was. Also present are Mark, an IT support teacher, and Adana, a girl in the class.

1. (4)

   

The reason this exchange was funny for the staff was that Brixton and Buxton are two very different places to get mixed up. Brixton, in South London, has a reputation for being a tough place, famous for the Brixton Riots in 1981, whereas Buxton, in Derbyshire, is a well-to-do and picturesque market town popular with tourists. In fact, Callum's misunderstanding goes deeper than this, as the Church Road he is talking about, famous for its gang violence and the grime artists Nines and Skrapz, is in fact in North London and not in Brixton at all. It is therefore plausible that Callum was especially aware at that point of how he was being perceived by others, and the self-correction was a fleeting example behind the scenes of this process.